AI-PS package notebook

setting working directory

knitr::opts_knit$set(root.dir = normalizePath("/Users/kanferg/Desktop/Gil_LabWork/Gil_packages/AIPS_Package/Example_images/sgPINKvsWT/"))

Parkin expressing cells treated with CCCP are loaded

frame<-load_image(Image_File_name="Image_1.tif",path="/Users/kanferg/Desktop/Gil_LabWork/Gil_packages/AIPS_Package/Example_images/sgPINKvsWT/",minmax_nrom=TRUE)

The variable object is a list of two channel Parkin-GFP in channel-1 and dapi staining for channel-2

display(abind(frame$CH1*5,frame$CH2,along = 1),method = "raster")

#par(mar = c(4, 4, .1, .1))
plot(frame$CH1*5)
plot(frame$CH2)

First the nucleus will be segmented using the Nucleus_segmentation function.

#par(mar = c(4, 4, .1, .1))
plot(frame$CH1*5)

plot(frame$CH2)

Than the GFP-Parkin will be segmented using the nuc_mask as the seed. The function retrieved a list composed of roi pixel coordination of the center of ecth cell and a table and two maks one for the cytosle and the other for the nuclues.

nuc_mask<-Nucleus_segmentation(x=frame$CH2,intens=1,filter_size=10,offset=0.04,opensize=3,
                               rmObjects_small=30,Use_watershed=TRUE,distmap_value=2,rm_outliers=TRUE,Out_P=0.95,
                               Masks_display=TRUE)
seg_disp = paintObjects(nuc_mask$mask_nuc, toRGB(frame$CH2), opac=c(1, 1),col=c("red",NA),thick=TRUE,closed=TRUE)
plot(seg_disp)

SVM model is created by picking cells which lost Parkin mitochondrial recruitment. These cells are labeled Positive. The function returns a feture_table contains 20 detures extracted from the segmented cells.

GFP_mask<-Cytosol_segmentation(x=frame$CH1,y=nuc_mask$mask_nuc,intens=5,filter_size=10,offset=0.1,mat_size_smooth=19,opensize=7,rmObjects_large=30000,Use_watershed=TRUE)

seg<-paintObjects(GFP_mask$mask_cyto,paintObjects(GFP_mask$mask_nuc,toRGB(frame$CH1*5),opac=c(1,1),col=c("red",NA)),opac=c(1,1),col=c("green",NA),thick=TRUE,closed=TRUE)  
plot(seg)

Picking phenotype exampled cells. The cells which Parkin mitochondrial translocation is abolished are labeled as postive cells for corresponding fetures marking.

Features_Table<-Feature_extract(x=frame$CH1,mask_cyto=GFP_mask$mask_cyto,label_class="Postive")
head(Features_Table$Ts.mix)

The data created will be assigned in to SVM algorithm and the accuracy of linaer karnel versus non-liner karnel is compered. All the tables shoud be save to the Table-training repository

plot(Table_training$seg_class_sel)

First we mesure acurrrecy for Linear Kerenel

non liner kernel

print(mean(acc))

[1] 0.9108896

Thea a PCA analysis for selecting features

print(mean(acc))

[1] 0.9146174

SVM model save and features selectopn acording to the PCA output

PCA_plot<-Fetures_PCA(Table_class_train=table,cartesian_lim_x=c(-15,10),cartesian_lim_y =c(-20,20),font_size=14)

Error in svd(x, nu = 0, nv = k) : infinite or missing values in 'x'
Error in svd(x, nu = 0, nv = k) : infinite or missing values in 'x'
Coordinate system already present. Adding new coordinate system, which will replace the existing one.

PCA_plot

For deploy svm model and creat a mask first Nucls and cytosole will be segmented and than the model will be deploy

##Deployment SVM

print(mean(model$ACC))

[1] 0.9183964

Now the model and the features selection will be used for the deploy and the genration of called cell mask

Features_Table<-Feature_extract(x=frame$CH1,mask_cyto=GFP_mask$mask_cyto,label_class="Postive")

Called from: Feature_extract(x = frame$CH1, mask_cyto = GFP_mask$mask_cyto, 
    label_class = "Postive")

Q

The selected cells which will get photoactivated

Target_mask<-maskGen_deploySVM(x=GFP_mask$mask_cyto,y=GFP_mask$mask_nuc,model=model$model,Table_class_train=Features_Table$Ts.mix,label_class="Postive",erode_mask=T,opensize=9,Sel=TRUE,Selected_Features=model$Selected_Features,TH=0)
display(frame$CH1,"raster")
celltext<-text(x= data.frame(Target_mask$table)["m.cx"][,1], y= data.frame(Target_mask$table)["m.cy"][,1], labels=data.frame(Target_mask$decision_values)[,1], col="yellow", cex = 0.8)